...：学习器在训练集上，实际预测输出与样本的真实输出的差异。可通过训练实验得到，往往和真实数据集有差异，所以该误差只能作为模型效果的参考。 泛化误差(generalization error)：学习器在新样本上，实际预测输出与样本的真实输出的差异。这是评估模型效果的标准。
基于2148个网页-相关网页
它基于结构风险最小化准则，目的是最小化泛化误差上界。
It operates on a principle, called structural risk minimization, which aims to minimize the upper bound on the expected generalization error.
最后，通过实验数据验证，系统的泛化结果误差小于6%。
Finally experimental data show that the system generalization error is less than 6%.
利用BP神经网络模型实现了对造纸废水处理过程的预测，平均相对误差为19%，表明网络泛化能力不是很好。
The effluent treatment process was predicted with this BP neural network model with the average relative error of 19%, which indicates that the generalization power of the network is not so desirable.
应用推荐